Poster Poster Program Therapy Physics

Utility of 18f-BPA PET Radiomics Analysis for Predicting Treatment Response In Boron Neutron Capture Therapy

Abstract

Purpose

The tumor-to-normal ratio of maximum uptake (Tmax/N) on 18F-BPA PET is widely used to guide decision-making regarding boron neutron capture therapy (BNCT). However, intratumoral heterogeneity of boron drug distribution may also influence clinical outcomes. This study aimed to evaluate the utility of 18F-BPA PET radiomics for predicting treatment response after BNCT.

Methods

We retrospectively analyzed 10 patients who underwent 18F-BPA PET/CT at National Cancer Center Hospital between August 2012 and June 2015, and subsequently received BNCT at the Kyoto University Research Reactor within 3 months. Best tumor response was assessed using RECIST v1.1 on follow-up images over 6 months after BNCT, and patients were classified as partial response (PR, n=5) or stable disease (SD, n=5). Using MIM Maestro® (MIM Software Inc., ver. 7.4.2), a radiation oncologist delineated tumors on CT images at 18F-BPA PET/CT, and the contours were transferred to PET images. A total of 107 radiomic features were extracted from intratumoral PET uptake using PyRadiomics. Predictive models for the tumor response were developed with leave-one-out cross-validation (LOOCV), with feature selection performed by Lasso regression using only the training data in each iteration. For each LOOCV model, the feature with the largest absolute coefficient was identified, and its selection frequency was summarized across iterations. A logistic regression model was then built using the single most frequently selected feature. Model performance was assessed by apparent AUC, and internal validation was performed using 1,000 stratified bootstrap resamples with optimism correction (Harrell’s method).

Results

The feature of 10Percentile was selected most frequently in LOOCV (8/10 iterations) and in bootstrap validation (379/1,000 resamples). The model achieved an apparent AUC of 0.88. Mean optimism was 0.18, resulting in an optimism-corrected AUC of 0.70 (95% CI, 0.40–1.0).

Conclusion

18F-BPA PET radiomics suggested that intratumoral low-uptake regions may be important for predicting treatment response to BNCT.

People

Kei Nakai, MD, PhDAuthors · Department of Radiation Oncology, Institute of Medicine, University of Tsukuba Minoru Suzuki, MD, PhDAuthors · Institute for Integrated Radiation and Nuclear Science, Kyoto University Yoshinori Sakurai, PhDAuthors · Institute for Integrated Radiation and Nuclear Science, Kyoto University Kouji Kunito, BScAuthors · Euro Meditech Co., LTD. Satoshi Nakamura, PhDCorrespondings · Division of Radiation Safety and Quality Assurance, National Cancer Center Hospital Hana Endo, BScAuthors · Division of Radiation Safety and Quality Assurance, National Cancer Center Hospital Teruhito Aihara, MD, PhDAuthors · Osaka Medical and Pharmaceutical University Nobuhiko Kamitani, MDAuthors · Division of Radiation Oncology, Kawasaki Medical School Teiji Nishio, PhDAuthors · Medical Physics Laboratory, Division of Health Science, Osaka University Graduate School of Medicine Tairo Kashihara, MD, PhDAuthors · Department of Radiation Oncology, National Cancer Center Hospital Hiroshi Igaki, MD, PhDAuthors · Department of Radiation Oncology, National Cancer Center Hospital Homare Murata, MScPresenting Author · Department of Radiological Technology, National Cancer Center Hospital Tomonori Goka, MScAuthors · Department of Radiological Technology, National Cancer Center Hospital Kimiteru Ito, MD, PhDAuthors · Department of Diagnostic Radiology, National Cancer Center Hospital Hiroaki KuriharaAuthors · Department of Diagnostic and Interventional Radiology, Kanagawa Cancer Center